I am designing a PoE powered development board with an ESP32 microcontroller module.

I am using the following circuit for the PoE power supply:

enter image description here

This is a direct copy of the PoE power circuit used by Olimex on their ESP32-POE boards.

Obviously, the behaviour I expect is a working ethernet connection regardless if the device is powered through the ethernet port or not and that the device will power on when connecting it to a PoE switch or injector.

The actual results I am seeing are the following:

  • When connecting to a non PoE switch with other power source: There is a reliable ethernet connection and everything works as expected.

  • When connecting to a PoE switch without other power source for the device: The device does not power on at all. Very low voltage (milli volts) measured from power traces on the PCB.

  • When connecting to a non PoE switch through a PoE injector: The device powers on and the activity LED on the switch and the device both turn on. The Link light blinks on the device but not the switch. The device does not receive an IP from the DHCP server and if I set a static IP, I still cannot establish any kind of connection to the device.

My thoughts on these symptoms:

Since the device does not turn on at all when connected to a PoE switch, It seems like the classification isn't working. It is however working for the PoE injector. When the device gets power from the ethernet port, there might be interference that disturbs the ethernet signals. Is this possible? I can provide a screenshot of the PCB traces if necessary.

Does anyone else have any insights on these symptoms and what the cause might be?

Edit 1: The data sheets for the ICs:

TPS2376DDAR-H: https://www.ti.com/lit/gpn/tps2375

TX4138: https://datasheet.lcsc.com/lcsc/1811141153_XDS-TX4138_C329267.pdf

Edit 2: The following schematic shows the ethernet connection: enter image description here

  • \$\begingroup\$ Please link the data sheets for the devices used. \$\endgroup\$
    – Andy aka
    Sep 22, 2023 at 12:34
  • \$\begingroup\$ @Andyaka I have updated the question \$\endgroup\$ Sep 22, 2023 at 12:51
  • \$\begingroup\$ "I am assuming then that the injector injects power constantly without trying to classify the device" - you're assuming wrong. \$\endgroup\$
    – Finbarr
    Sep 22, 2023 at 13:00
  • \$\begingroup\$ @Finbarr Maybe I am. I do not have the equipment I need to verify it at he moment but I will check as soon as I can. Do you have any other thoughts on why the injector powers it on and and the switch doesn't? They are both IEEE 802.3af compliant and working with other PoE powered devices. \$\endgroup\$ Sep 22, 2023 at 13:06
  • \$\begingroup\$ There's no schematics for the Ethernet connector or data path. Just the PoE module input and output. \$\endgroup\$
    – Justme
    Sep 22, 2023 at 13:33

1 Answer 1


I'm a bit concerned about your voltage feedback resistors and compensation capacitor on the TX4138. The data sheet recommends these values: -

enter image description here

And, although you appear to be using a 12 pF capacitor, your feedback resistors appear to be ten times lower in value than the recommended circuit above. This could cause problems despite you saying it matches an already proven circuit: -

enter image description here

Also these are no details in the data sheet about the internal pull-up on the enable input. This has to be strong enough to suit the open-drain output of the TI chip: -

enter image description here

Image from here

Here's your circuit in the area I'm talking about: -

enter image description here

This is a direct copy of the PoE power circuit used by Olimex on their ESP32-POE boards.

Are you sure about that? Maybe there's an external pull-up that you missed? Mind you, the feedback resistors and capacitor are a significant worry.

  • \$\begingroup\$ Thank you for pointing out the feedback resistors, this might absolutely be a problem source. I also agree that there should probably be a pull-up on the PG-pin although I do not think this is why the device does not power on. When I say that I measured the voltage on the power traces, I mean the power traces coming from the ethernet port (between spare1 and spare2 in the diagram) which would be unaffected by the PG signal. I will investigate and get back to you. \$\endgroup\$ Sep 22, 2023 at 13:21
  • 1
    \$\begingroup\$ The other thing about PG is that if it enables the converter chip too quickly you can get in a bit of a pickle where the inrush monitor circuit thinks it has reached an OK value and asserts PG and enable the buck. This can cause the buck to draw too high a current and exceed the inrush limit and you get a situation where it turns on and off rapidly. You could try adding a 100 nF capacitor from PG to ground to alleviate this. You are trying to avoid the input supply sagging below the UVLO threshold. \$\endgroup\$
    – Andy aka
    Sep 22, 2023 at 13:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.